lelu
Integrations

Anthropic

Authorize Claude tool use through Lelu before execution. Works with the Anthropic SDK, Claude API tool_use blocks, and the Claude Agent SDK.

Installation

terminal
npm install lelu-agent-auth @anthropic-ai/sdk

Tool use with Claude

Try transfer_funds in sandbox

Claude returns tool_use blocks when it wants to call a tool. Pass the tool name to Lelu before dispatching.

agent.ts
import Anthropic from "@anthropic-ai/sdk";
import { createClient } from "lelu-agent-auth";

const claude = new Anthropic();
const lelu = createClient({ apiKey: process.env.LELU_API_KEY! });

const response = await claude.messages.create({
  model: "claude-opus-4-5",
  max_tokens: 1024,
  tools,
  messages,
});

for (const block of response.content) {
  if (block.type !== "tool_use") continue;

  const decision = await lelu.authorize({
    tool: block.name,
    context: JSON.stringify(block.input),
  });

  if (decision.decision === "deny") {
    console.error(`Blocked: ${block.name} — ${decision.reason}`);
    continue;
  }

  if (decision.decision === "human_review") {
    await requestApproval(decision.requestId);
  }

  await executeTool(block.name, block.input);
}

Claude Agent SDK

Use a pre-tool-call hook to intercept every tool invocation automatically.

agent.ts
import { Agent } from "@anthropic-ai/agent-sdk";
import { createClient } from "lelu-agent-auth";

const lelu = createClient({ apiKey: process.env.LELU_API_KEY! });

const agent = new Agent({
  model: "claude-opus-4-5",
  tools: [searchTool, writeTool, deleteRecordsTool],
  hooks: {
    beforeToolCall: async ({ tool, input }) => {
      const decision = await lelu.authorize({ tool: tool.name });
      if (decision.decision === "deny") {
        return { abort: true, reason: decision.reason };
      }
    },
  },
});

Environment variables

.env
ANTHROPIC_API_KEY=sk-ant-...
LELU_API_KEY=lelu_sk_...
Previous: OpenAINext: LangChain